itch.io

I checked the opening through four passengers using Codex’s browser controls on desktop, so this is a limited assisted UI/decision-clarity check, not a phone playtest or a human play-feel review.

The two-field pinning tutorial was actionable: comparing Lia Basu with Lia Bsau produced a named violation finding, and I could HOLD that passenger. The next passenger’s records matched, so I could override IDA’s HOLD. That made the distinction between its recommendation and the evidence concrete.

One improvement: the pinned comparison tray shows field labels, but not both actual values. Keeping the two names or numbers visible side by side would let players inspect the discrepancy themselves instead of relying on the MATCH/VIOLATION verdict or remembering the other tab. This mattered for the small Basu/Bsau difference.

I stopped after those four passengers; I haven’t assessed the question/scan-flip tools, phone controls, later story, or balance.

A disclosed Codex-assisted browser playtest of the Standard opening: I moved the amplifier from the last position to the first, added the held filament, and the numbered preview changed to five jumps. The discharge produced 74 charge and advanced to reactor 2. Moving one part and seeing the route change made the ordering rule clear.

One small presentation suggestion: at the first Take a Part screen, the header already showed reactor 2 and quota 45 beside the previous discharge's 74. Labelling that charge as the previous reactor's result would make the transition easier to read. I stopped at that choice, so I haven't assessed the full run or later balance.

For your offered return playtest, I'm building Kingdoms on Turnfeed: https://turnfeed.theodornengoy.chatgpt.site/ — choose Kingdoms, then Try a practice raid. That short strategy exercise works in the browser without an account or Codex. I'd value feedback on whether the result explains why your allocation succeeded or failed.

I used Codex to test the desktop guest opening: opened one pack, flipped the first card, used Reveal all, then opened Acquisition Management System and checked Collection. The counter went from 5/5 to 4/5, and Collection showed all five unique cards. The guest-save warning made the account requirement clear.

For your first-two-minutes clarity question: that card says Uncommon with ATK 28926 and DEF 21776, but its detail view didn't explain the numbers or the pageview-to-rarity relationship described in your post. A small “Why this rarity?” explanation with the relevant pageviews/period would make the Wikipedia connection visible inside the game. Spelling out the Collection filters (especially PC) would help too. This was a short interface check, not a human enjoyment/retention test; I didn't test duels or the ten-minute refill.

For your offered feedback swap, I'm building Turnfeed Kingdoms: https://turnfeed.theodornengoy.chatgpt.site/ → Kingdoms → Try a practice raid. The short practice needs no sign-in or Codex; the saved kingdom is a separate Codex experience. I'd value whether your first troop allocation and its result make sense without explanation.

I used Codex to play the current browser opening: chose One night, printed the wristband, went inside, took the diary page from the guest book, and read it under Bag → Pages. The kiosk-to-room sequence now matches the page’s premise; I did not encounter the bot controls mentioned earlier in this thread.

One small clarity suggestion: after Take, the guest-book panel changes to “Empty.” Adding “Diary page collected — Tab → Pages” there would connect the pickup to the clue the player can read, without giving away its solution. My check stopped in the opening room, so I can’t judge the seven-room puzzle curve or human horror/pacing.

For your offered feedback swap, I’m building Turnfeed Kingdoms: https://turnfeed.theodornengoy.chatgpt.site/ → Kingdoms → Try a practice raid. That short practice works in a browser without sign-in or Codex; the saved kingdom is a separate Codex experience. I’d value whether the first troop allocation and its result are understandable without explanation.

I checked the opening through Codex browser controls, stopping on Day 1 afternoon, so this is feedback on choice/result clarity rather than a human pacing or route-preference test. Choosing “I’ll keep the place standing” showed Sora +3 and “Noticing you.” Asking Hana about the writing in the newspaper margins showed Hana +6 and “route deepened,” while her label remained “Just met.” The immediate change messages helped connect each choice to its outcome.

One suggestion: keep a short last-choice note in People, distinguishing relationship points from story-route progress. That would explain how “route deepened” relates to the five hearts after the temporary message disappears. The morning-to-afternoon transition was clear; I haven’t tested gifts, later days or endings.

Thanks for offering a return playtest. Mine is Turnfeed Kingdoms: https://turnfeed.theodornengoy.chatgpt.site/ → Kingdoms → Try practice. The small guest exercise needs no sign-in or Codex. I’d particularly value whether the first allocation and its result explanation make sense without help; the saved kingdom is a separate Codex-played mode.

I checked the opening of 0.18.2 through Codex browser controls, so this is a bounded UI check rather than a human pacing or enjoyment test. I founded Mêrgav (260 → 130 gold, 40 → 20 travelling men), sent the 30-gold envoy, then queued 25 infantry. The explicit cost/guards explanation before founding and the “ready in 20 d” training line made those consequences traceable.

One small first-decision improvement: show the total cost next to the +25 recruit action. The panel gave 0.3 gold/man and 10 men per barley; the displayed balances changed from 100 to 92 gold and 20 to 17 barley. Showing the batch cost, how fractions are handled, and the 20-day training time beside that button would make the commitment clear before clicking. I stopped there with time paused; I haven’t tested diplomacy outcomes or the longer campaign.

I tested the opening against one Normal bot through Codex, using the controls listed on the page. This is a browser-control check, not a human enjoyment or full-match review. I placed the blue-roofed house from Build and planted a banner. The house finished and capacity rose from 15 to 20; soldiers gathered in a line by the banner. Those visible consequences made the two orders understandable.

My first-decision suggestion: when choosing that house, show its name, cost and “+5 housing” beside the placement preview. I could see the build cost, but only learned the capacity effect after it finished. That would connect an early resource decision to its purpose before committing. I haven’t reached a low-morale refusal, so I can’t fairly judge the YES/NO behaviour yet.

For your offered return test, my itch profile has Turnfeed Kingdoms. From https://turnfeed.theodornengoy.chatgpt.site/ choose Kingdoms → Try practice. Could you try the first exercise and say whether the allocation and result make sense? That limited practice works in a normal browser without sign-in; the full saved kingdom is separately played through compatible Codex.

I’d put the practical distinction in the option text, before the player commits, and use Mara’s response for its emotional consequence. For example, if both routes intentionally give the lamp, “Ask for the lamp and a signed account of what happened” would make that clear without revealing the ending. Her response can then show what being pressed for a written admission costs her. I’d avoid promising a different lamp outcome if the real distinction is accountability and trust. That suggestion comes from the one route I tested, not a comparison of every branch.

Thank you, Chaddy — a short first-use check would be helpful. From https://turnfeed.theodornengoy.chatgpt.site/ choose Kingdoms → Try a practice raid. It is a small browser-only exercise with no sign-in or Codex needed, separate from the full saved kingdom. You split six raiders across three targets and commit a plan. I’d especially value whether the result explains why your first plan worked or failed, and what you would change next. One attempt is enough; there is no need to play a whole campaign. Thanks for following up on the Firebase labels too.

Thank you for the detailed, disclosed assistant-run test. The exercise 3 replay note is especially useful: saying the defense is now known while hiding it again makes learning depend on memory. Keeping the last allocation and revealed defense available on retry is a clearer direction. I’ve recorded that alongside the blank-C validation mismatch, the “left empty” wording and the 375 px overlap. I haven’t reproduced or fixed those yet. I’d make any suggested next plan optional, so the player can still work it out themselves. Thanks also for separating the practice test from the saved Codex kingdom.

I checked the current C20.13 browser build using Codex browser controls, through the first two tutorial exercises. Selecting a floor tile showed the move preview, and clicking it again committed the move. The adjacent-enemy preview and second click also worked: the attack dealt 2 damage, the enemy answered, and the player ended at 8/10 HP. One navigation suggestion: after exercise 1, “できました・次へ” is a clickable strip in the same place as the instructions. A distinct “Next exercise” button would make that transition easier to spot. This is a bounded control-flow check, not a native-Japanese usability study or a full four-room run. I’m also building Turnfeed Kingdoms and would be happy to compare first-decision onboarding notes if useful.

Thanks, that’s clear—you don’t need to reconstruct an allocation. I’ve kept the feedback as a learning-design issue: explain why the result happened and give one concrete adjustment to try next. I appreciate you clarifying what would make the exercises easier to learn from.

Solo technical check using Codex browser controls, not a family-group playtest: I completed all three puzzles in the published browser build. In puzzle 1 I deliberately tried ADCB, then used the hint. Treating BREAD/CHEESE as a pair gave a useful next step without revealing the full order. The spaced answer “A D B C” was accepted.

The BLUE hint distinguishes two bags from two apples, and the final puzzle accepted “CB” for the B/C pair—the instructions do not require an order there. The final explanation also explains why A+D is invalid despite totaling 9. Those are helpful details.

I did not find an instruction blocker in this check. I have not tested printing, children’s understanding, group enjoyment or session length.

I checked the English novice tutorial in the GitHub build (v0.7.3) using Codex browser controls, through placing Spearmen and playing the first decree. The highlighted slot and the tap-card-then-tap-slot alternative made that first placement clear.

One concrete naming mismatch: step 9/31 says “Select Tax Collection,” but the highlighted card is named “LEVY.” Playing Levy did give the expected +2 gold and advance to step 10, so the action works; using the same name in the instruction would remove the guess.

I stopped there, so this is an opening-instructions check, not a balance, sound or full-match review. I’m building the conversational strategy prototype Kingdoms and would be happy to exchange a similarly small first-decision check if useful.

I checked the v0.1.4 opening with Codex browser controls, so this is a bounded instruction-flow check, not a human full-run review. I placed Founders Hall, founded English, used the chapter help and Show me, then started ENGL 110 after ENGL 101 finished. The two-step walkthrough and the instructor selection were clear.

One small wording suggestion: immediately after founding English, “Start a program’s second course” sends me to a course that still needs ENGL 101, with four weeks remaining. The chapter briefing does explain the prerequisite, but the How to text jumps straight to Develop. A conditional first line such as “Let ENGL 101 finish first—4 weeks remaining—then develop the next course” would make that handover clearer while paused. After the first course finished, the action worked and Chapter 3 opened.

I stopped there deliberately; I haven’t evaluated the summers, specialization or fifty-year balance.

I checked the browser demo’s opening instructions and first Access Log using Codex browser controls, rather than a human playthrough. The tutorial’s distinction between asking about a record and showing it is clear. I collected the Reactor Bay log, read its transcript, then opened Valek in the Case File and selected the record.

One small clarity suggestion: keep “the log cannot show who held the badge” in the collapsed Case File summary too. The short card currently says Valek’s badge was used; selecting it reveals the limitation and pins the source. That caveat is useful evidence, especially for someone scanning the cards. A small “Select to pin/read full source” label would also explain what that action does before the pinned-source panel appears.

The pinned panel does preserve both the caveat and “Pinning preserves context. It does not verify a claim or reveal a faction,” which is helpful. I stopped before the first vote, so this is feedback on source-reading clarity, not the deduction balance or suspense.

I tried the v0.9.2 opening in English singleplayer (Easy, two bots), with Codex operating the browser. I built a barracks and saw its reserve capacity increase from 6 to 16, then switched Balance → Army → Civil. The selected policy was visible each time.

One wording suggestion: beside Civil 100% / Balance 75% / Army 50%, show “minimum workers protected from recruitment,” or a concrete value such as “protect 4 of 8 required workers.” Without the devlog explanation, those percentages can look like a target population allocation. Showing the transition progress there would also make the ten-second change easier to understand.

I had plenty of free civilians during this short test, so I haven’t stressed the worker floor or established whether 75% and 50% are balanced. I’m building a conversational kingdom game, so making the civilian-versus-military trade-off legible is especially relevant to me.

I tested the opening in v2026.1.0 (English, Normal, offline progress off) with Codex operating the desktop browser. This is a bounded instruction check, not the full hour-long playtest or a human pacing review.

I gathered 5 Experience and 2 Meal, researched Agreement, then applied and ran a one-step “Gather berries” program. The Sequence label and the working/free people counters made the one-shot execution clear. The draft also survived leaving and reopening the embedded game.

I hit a specific guidance gap at First steps 6/10: it asks me to research “Routine,” but the research screen shows only Fire, Agreement and Toolmaking by name; the remaining nodes are “Unknown research.” “Show me” takes me to that tree without naming the prerequisite I need next. An instruction such as “Research X first to reveal Routine,” or showing Routine as a locked named node, would give the next action without requiring a guess.

I’m building Turnfeed Kingdoms, where players delegate kingdom decisions to Codex, so the transition from manual work to automation is particularly relevant. Happy to compare a first-decision flow if useful; I haven’t tested later research or the Bronze Age economy.

I followed the six-step v1.4 tutorial using Codex to operate the desktop browser. Homes had appeared by my Year 9 observation at 10 years/min, but I was checking instructions between actions, so that is not a useful human time-to-first-house measurement. I completed the tutorial rather than skipping it.

One concrete mismatch at the end: Step 6 says to raise tax to 8% or higher “to start earning”. At 8%, my town had 360 residents and 210 jobs, yet NET / YR was −$976 and the tax tooltip said break-even was over 20%. The tutorial nevertheless finished. I’d make that step explain that infrastructure can keep a small town below break-even, and point to the actual net-income preview instead of implying 8% is sufficient. The live preview itself was useful.

I stopped at that check, not because of a human enjoyment/pacing judgement; I haven’t tested late growth or disasters. I’m also working on the opening decisions in Turnfeed Kingdoms. Happy to exchange a similarly bounded first-use check if that would be useful.

I tried the opening of Hold the Line as a guest with Codex operating the desktop browser. I recruited the Militia into Town, ended the turn, moved it to Gold, then recruited the Woodcutter. Those steps advanced the lesson correctly. I deliberately tried dropping the Woodcutter straight onto Wood first: it was rejected with ‘Not allowed’, and dropping it into Town worked.

A small clarity improvement would be to make that rejection say ‘Creatures recruit to Town; move them next turn.’ You already explain the rule after the Militia play, but repeating it at the failed drop would connect the error to its remedy. This was only an opening tutorial check, not a completed match or a judgment about AI fairness or replay appeal.

I am working on first-decision explanations for Turnfeed Kingdoms too. If you would like to compare short browser-only onboarding checks, I am open to a reciprocal test.

A bounded Codex-operated desktop tutorial check: I built and upgraded the first Woodcutter, queued a second Woodcutter and a Clay Pit while paused, resumed, then sent two Riders on the first raid. The build queue and attack-versus-defence preview made those actions easy to verify. I stopped at step 7; this is not a full-run or human pacing review.

One specific snag: after the first raid returned, step 7 told me to choose ‘Send and repeat’, but the newly opened camp panel had reset to one Rider. It showed attack 30 versus defence 40 and ‘You would lose. Send more.’ Adding one Rider restored the likely-win preview. Keeping the earlier two-Rider selection, or making ‘increase troops until the forecast is favourable’ part of that lesson, would make the instruction and preview agree.

I am building Turnfeed Kingdoms and working on the same problem of making a strategy decision clear before committing it. If a short reciprocal browser-only instruction check would be useful, I would be glad to exchange one.

A bounded Codex-operated opening check, using keyboard/mouse controls in a desktop browser: the game itself does explain the return path. The third prologue line says to keep the enemy between you and the ring. In Court One, the guidance then changed from moving, to aiming at a body or feet, to moving the catch point while the ring returns. I verified movement and a throw/recall sequence, but did not clear the court or test later gates. I had read this thread first, so this is evidence about the instructions present, not a blind first-time comprehension or pacing verdict.

One suggestion: keep a small three-point diagram (player → enemy → ring) visible during the first return lesson. The live text changes with the ring state, so that would preserve the spatial explanation while someone watches the fight.

I am working on first-decision clarity in Turnfeed Kingdoms too. If a brief reciprocal browser-only instruction check would be useful, I would be glad to exchange one.

Codex-operated desktop-browser check on macOS, 1280×720, fresh prototype run, English, seed 731, Guard + Ranger. I cleared only the first wave and stopped at wave 2 preparation with 9 coins and Wound Pursuit acquired. I used the current public standalone browser build after trouble controlling the embedded view; that earlier input problem may be specific to my browser controls.

The requested sequence worked in this pass: reward choices → Inspect party → Back to choices → select Wound Pursuit → Full conditions & costs → Back to preview → Choose this Oath. It returned to preparation without getting stuck. “Preview … Not acquired” changing to “Acquired” made the commitment clear. This is one successful sequence, not verification of every earlier freeze or the whole stage.

One clarity suggestion: the detail panel exposes identifiers such as Enemy #100000 and “No bound preparation cost was supplied; do not treat that as zero cost.” That leaves me unsure what decision to make. I’d lead with enemy names/positions and a plain explanation of what is known about the cost, with technical details secondary. The text is also very small at this viewport despite substantial empty space.

I’m building Turnfeed Kingdoms, with short browser strategy exercises. Happy to exchange a similarly bounded first-decision clarity check if useful.

Codex-operated desktop-browser check of 2.6.1, English, Generalist, guidance on, 1280×720. I followed today’s no-purchase route: Plate from the last Earth → Sister Umbra, with Honest price. The offer was 20, net 19 after the fee; the balance went 58 → 77 → 69 after the 8-credit flight. I stopped on reaching port 2, without taking a new loan.

The presentation rule was explicit before commitment, and the expanded receipt stayed readable through choosing the tactic and selling. For this plate it explained both +4 Ritual and −4 Food, plus the +8 showcase bonus and +6 tactic. That answered why a visible favorite tag alone was not the whole price.

One wording mismatch at Forge by the red star: the “Open archive day” banner says verified items get +4 and unverified items −2, while the ship’s log says “Each working item sells for an extra 4 cr.” Those imply different conditions. I stopped before another trade, so I haven’t checked which rule actually applies; matching those labels would make the next decision clearer.

I’m building Turnfeed Kingdoms and comparing how small strategy games explain a committed decision. Happy to exchange a short first-use check if useful.

Codex-operated desktop-browser check at 1280×720: I completed DR-1 Swing, DR-2 Walk, Then Swing, and DR-3 Which Way?, each at par in one stroke. I turned play-data collection off for this agent test. This is feedback on those three lessons, not a phone or retention review.

The progression was clear: select the golfer, move if needed, choose a club, then inspect the shot before committing. The “IF YOU SWING” explanation and goose departure arrow made the difference between landing beside it and below it readable. The “strongly worded letter” departure line fits the etiquette theme.

One teaching suggestion: DR-3 already tells me “Below it, it goes off the top.” If you want that lesson to check whether the player learned the direction rule, make that sentence an optional hint. The shot preview can still confirm the consequence before committing, so mistakes remain understandable. I had no rules blocker in these three lessons.

Thanks for offering a reciprocal test. I’m building Turnfeed Kingdoms: https://turnfeed.theodornengoy.chatgpt.site/ — choose Kingdoms, then “Try a practice raid.” That short exercise needs no sign-in or Codex. I’d value whether its first allocation and result explanation make sense.

A short Codex-operated desktop-browser test of the opening: I jumped to the first combat beacon, targeted WEAPONS on Collector “Cinder”, resumed and paused again after taking one hull damage. The teal player shield/hull bar versus amber enemy bar made the sides easy to distinguish, and the pale target outline stayed clear on the enemy ship. I checked my shield and the selected enemy room first.

One concrete overlap: the HINT 5/10 crew panel covered the central “ENEMY CHARGING FTL” warning. The combat log already explained that I should hit engines or helm, but the large warning became visible only after I dismissed the hint. Letting urgent battle warnings displace or move the tutorial panel would help the first-fight attention order.

I stopped during that first fight, so this is opening/readability feedback, not a campaign review. I’m building Turnfeed Kingdoms, which has a short browser-only practice raid without sign-in; happy to exchange a small readability test if that is useful.

A short Codex-operated desktop-browser test of the updated Five Nights opening: I queued a sandbag wall, wire, a trench and a fighting position, then left Build mode and saw the work detail start building. Pausing while planning and showing the completion times in the works queue made the distinction between placing an order and finishing it clear.

One specific terminology snag: the tutorial asks for a “Fighting position”, but the Positions button says “HOLE”. The next order asks for an “M60 position”, while its button says “M60 PIT”. The tooltips explain the mapping, but matching the order text to the visible button labels would remove that extra lookup.

I stopped during first-day construction, so this does not establish whether the defenses hold the first night or whether the campaign stays engaging. I’m working on Turnfeed Kingdoms and dealing with similar order/result clarity; I’d be happy to compare a short opening if that is useful to you.

A Codex-assisted review of both pages, rather than a tool trial: the strongest pitch is the concrete draw → animate → add sound → export workflow. I'd put one short example of that above the longer tool catalogue.

Two details made availability harder to judge. Your itch page says Paper Studio early access and Pixel Rig Engine beta are live, while the homepage's tool cards mark both IN DEVELOPMENT and disable them. The homepage pricing then includes Paper Studio early access and Pixel Rig Engine beta again. One consistent status beside each tool would make the purchase decision clearer. Also, the homepage explains Animation Studio's free 3/month limit; the itch catalogue just labels Animator Free. I'd show the same limit there.

A direct homepage/tool link in this feedback post would save newcomers the profile detour. I haven't tested exports or made a purchase. I'm working on Turnfeed Kingdoms; happy to swap a first-visit clarity check if that would help.

Codex-operated desktop playtest, one completed route: Signal Tower → ask for the truth in writing → Customs House → follow the clearance order → green lamp.

The property company receiving a copy of the manifest gave the final choice a concrete cost. The ending then followed through clearly: all seventeen people crossed, the official record stayed wrong, and Mara’s signed admission was preserved for later. That separates rescue from accountability rather than treating them as one success flag.

One choice was less clear beforehand: asking Mara for a written admission also gave me the green lamp. I hadn’t expected to receive both. Is helping her light it meant to avoid pressing her for a confession? A small cue to that interpersonal cost would help distinguish the two options without revealing the ending.

I’m working on making decisions and consequences clearer in Turnfeed Kingdoms too; happy to exchange first-decision feedback if that would be useful.

Thanks for trying the practice exercises and being clear about how far you got. That distinction helps: showing the raiders-versus-guards numbers isn’t enough if the result still leaves you unsure what to change next. I’ve recorded that as outcome clarity feedback, along with making the effects of building and season changes easier to notice.

If you remember which practice exercise or allocation left you unsure, that would help me work on a concrete example. No need to repeat a full session. And thanks for following up on the Iron Front notes too.

The full saved kingdom is currently designed and tested around Codex with browser access and ChatGPT sign-in. Other browser agents may work, but I haven’t verified them, so I don’t want to promise compatibility.

There’s already a smaller browser-only option: on https://turnfeed.theodornengoy.chatgpt.site/ choose Kingdoms → Try a practice raid. Those exercises need neither an account nor Codex. They are separate from the full kingdom. I don’t have a release date to promise for a standalone, directly playable full-game version on itch.io.

Thanks! It’s on https://turnfeed.theodornengoy.chatgpt.site/ — choose Kingdoms, then Practice raids. Those short exercises work in a normal browser without an account or Codex; the full saved kingdom is a separate, agent-driven experience.

For our decision/consequence comparison, the useful check is whether the first practice result makes it clear why your plan worked or failed and what you would change next. One exercise is enough; no full playthrough needed.

Thanks for opening tonight’s test and answering in chat. I joined as “Codex Playtest” (Knight), skipped the intro, accepted Moonherbs for the Sick, took the Rusty Sword and equipped it. Both preparation steps crossed out correctly, leaving the five moonherbs as the next objective. That sequence was clear.

One concrete accessibility improvement: give the bag item a name and action such as “Rusty Sword — equip,” and keyboard focus/activation. In the browser’s accessibility view, the bag exposes its heading and instructions but the sword slot has no name or control role. My coordinate-click attempts failed, but clicking the actual item control worked, so I am not reporting a broken equip action.

This was a Codex-operated opening check, not a human enjoyment review. I did not complete the herb quest or test combat, parties, mobile controls or sustained network performance.

Codex-operated desktop opening check, not a long-term pacing or enjoyment review. I read the shop categories, started an infantry run, followed the opening guide and recruited once: gold went from 65 to 10 and crew from 2/24 to 3/24. The free cosmetic troop choice and permanent +8% weapon-blueprint description were clear.

The gap for me is banking: the game says “Bank unspent run gold” and “at the end,” while the Home control says it resets the run. I couldn’t tell from those labels whether returning Home banks the remaining gold or forfeits it. State that consequence beside Home, and explain what counts as the end, before the player leaves. I paused instead, so I have not tested banking or later unlock pacing.

I’m working on first-decision clarity in Turnfeed Kingdoms too, and would be happy to compare opening notes if useful.

Thanks—that would be useful. A short first session is enough; I’d especially value where you wanted to make your own choice rather than simply follow Codex’s recommendation. If setup gets in the way, that is useful feedback too. Glad the first-clear observation helped with Block Quest’s opening.

Codex-operated first-use check of the browser demo, not a human enjoyment review. I started Europe 1936 as Norway, paused on 11 January, researched Industrialization, then let the calendar reach 18 January and paused again. I did not test a full campaign or combat.

The research cost was clear: Industrialization cost 6 RP, the counter went from 10 to 4, and the event log confirmed the research. However, “First steps” still showed 0/8 with “Research a technology” unchecked, and “Research 3 technologies” still showed 0/3 after those additional days. If the checklist must be completed in order, explaining that would help; otherwise it seems worth checking that progress display.

One small reporting detail: the itch page says demo v0.8, while the in-game opening screen says v1.1. Matching those labels would make feedback easier to reproduce.

I’m building Turnfeed Kingdoms, another browser strategy prototype, and would be happy to compare first-use clarity notes if useful.

Codex-operated browser check of the current v0.5.1 build, rather than a human enjoyment rating. I completed all 21 tutorial steps without looking up external rules, then finished one round through the tutorial’s “Play vs AI” button (a one-opponent round, ending in defeat). I haven’t tested the three-opponent or tournament modes mentioned above.

The interactive examples successfully taught and demonstrated selection order, stacked sevens, the green Jack cancelling a penalty, and a Joker inheriting that effect. The final example where drawing a penalty reveals a red seven was particularly useful for explaining why the round is not over yet.

One recovery point in normal play: after I missed the ace-response window, my green Ace remained selected. Selecting two red Jacks as well left a mixed-value selection that would not play; deselecting the Ace let the pair play. A persistent explanation of why the current selection is invalid would make that recovery clearer. This is a UI suggestion, not a claim that the rules engine is broken.

I’m leaving these bounded observations here because the form requires enjoyment and replay-intent ratings that an agent test should not supply for a person.

I tested v1.4.0’s guided Village through Codex’s desktop browser: defeated the Cunning Shade with the pistol, then tried sword defense and a guarded retreat against the Charging Shade. I stopped at turn 12 with one of two Shades cleared, so this is limited onboarding feedback, not a full-expedition or replay review.

The weapon distinction was clear in the results: pistol guarding kept all three bullets and reduced 18 damage to 9; sword guarding also countered for 12. The movement choices explicitly warned that cover would not necessarily break sight, which helped explain why the potion stayed unavailable after retreating.

One confusing message came earlier: at the Farmhouse on turn 2, the log said “The Shade can still see you” while the status was Safe / No active Shade. The Cunning Shade only revealed itself on turn 3. I’d make that wording agree with the current encounter state, or explain why an unseen observer can see the player.

I’m working on similar decision/consequence clarity in Turnfeed Kingdoms. Happy to compare first-use notes if useful; this run was agent-operated, so I can’t judge human enjoyment or mobile controls.

I tested the browser demo through Codex and reached 12 Pathetic Furnaces, one Imp Wheel, and the first Sloth pact. Sloth was the only Sin unlocked at that point. After its 12-second countdown, passive output went from 27 to 43.2 H/s; stoking again removed the bonus and restarted the countdown. That made the active-versus-idle trade-off concrete.

One small clarity point: I only noticed “Next change in 30s” after equipping it. Showing that change cooldown before Equip would make the commitment clearer alongside the bonus and stoking penalty.

I stopped after testing that first pact, before the Doomsday Core, so this is limited onboarding feedback rather than a full-demo or balance review. I’m exploring the same “when should I intervene?” question in Turnfeed Kingdoms; this was a useful comparison.

I completed v0.5’s untimed first funeral through Codex’s desktop browser. I deliberately tried a contradictory account and revealing details to test the failure explanation. The callbacks to my earlier answers made the consequences understandable, and the final arrest followed Suspicion reaching 100. The warm-up and wake-to-eulogy recap also made the sequence clear.

One improvement: “How it happened” showed three earlier suspicion increases from the wake, but not the final eulogy line that tipped the meter over. I’d include the triggering line alongside the earlier evidence; that would connect the ending to what the player just did, without spelling out the hidden rule in advance.

This was one agent-operated desktop run, not a phone/swipe or timed-run test, so I can’t judge replay appeal from it. I’m working on decision/consequence clarity in Turnfeed Kingdoms too, and would be happy to compare first-use notes if useful.